Issues when combining two datasets

I have two datasets from adjacent areas, that were collected separately but have sufficient overlap that I believe they should be able to map as a single task.
Dataset 1 has ground control points, and processes cleanly, with good repeatability to historic data.
Dataset 2 does not have control points, as it was collected as an afterthought, but extends from one edge of dataset one.
When processed separately, each dataset looks good. Dataset 2 has some height issues due to the lack of GCPs but correct general trend is there.
When I try to process together, the resulting orthophoto, surface model and point cloud, only seem to cover Dataset 1. I thought at first that the images from dataset 2 were not being used correctly, but when I check their positions, they show up on the map, but in the wrong place, with no data below them!?

I’ve uploaded some screenshots of what I’m experiencing along with my log here…
If I can add any more information that would help to see what’s going on please let me know.

1 Like

Sorry to bump this but would really appreciate some help if someone can point me in the right direction for any additional settings to try/info needed/logs etc etc.

I’m still learning the ropes here myself, but I could at least try to reproduce your issue. Could you make the dataset available somewhere?

Just a thought, have you tried combining the datasets without the GCPs?

What drone/sensor was used for image collection? Was it the same settings for both datasets?


1 Like

Sure, It’s pretty large but I’ll try and upload and send a link.
I did try the data without GCPs and got the same result. (I thought the distance from the GCP’s may have been an issue).
Data were all acquired using Drone Deploy, so settings should be similar. The one flight was done with a Mavic Pro, while the other was a Mavic Pro platinum. As I understand it they should have the same camera/sensor setup, and it’s only the props/ESCs/motors that changed between the two models.

1 Like

It is possible to mix sensors, so even if there are small differences, that’s fine, but I agree those should have the same sensor.

1 Like

Woody, you have a PM! If anyone else would like to try the dataset please let me know and I’ll drop you a PM with a link. (It’s a commercial/research dataset, so prefer for it not to be publicly available at this stage).
The dataset has some of it’s own issues, motion blur etc, but as they have reconstructed separately, I don’t think that should be an issue here.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.